Academia.edu no longer supports Internet Explorer.
To browse Academia.edu and the wider internet faster and more securely, please take a few seconds to upgrade your browser.
The role of verb repetition in cumulative structural priming in comprehension
The role of verb repetition in cumulative structural priming in comprehension
The role of verb repetition in cumulative structural priming in comprehension
The role of verb repetition in cumulative structural priming in comprehension
The role of verb repetition in cumulative structural priming in comprehension
T. Florian Jaeger2016, JEP: LMC
Recently processed syntactic information is likely to play a fundamental role in online sentence comprehension. For example, there is now a good deal of evidence that the processing of a syntactic structure (the target) is facilitated if the same structure was processed on the immediately preceding trial (the prime), a phenomenon known as structural priming. However, compared to structural priming in production, structural priming in comprehension remains relatively understudied. We investigate an aspect of structural priming in comprehension that is comparatively well understood in production, but has received little attention in comprehension: the cumulative effect of structural primes on subsequently processed sentences. We further ask whether this effect is modulated by lexical overlap between preceding primes and the target. In three self-paced reading experiments, we find that structural priming effects in comprehension are cumulative and of similar magnitude both with and without lexical overlap. We discuss the relevance of our results to questions about the relationship between recent experience and online language processing.
Related Papers
When we read or listen to language, we are faced with the challenge of inferring intended messages from noisy input. This challenge is exacerbated by considerable variability between and within speakers. Focusing on syntactic processing (parsing), we test the hypothesis that language comprehenders rapidly adapt to the syntactic statistics of novel linguistic environments (e.g., speakers or genres). Two self-paced reading experiments investigate changes in readers’ syntactic expectations based on repeated exposure to sentences with temporary syntactic ambiguities (so-called ‘‘garden path sentences’’). These sentences typically lead to a clear expectation violation signature when the temporary ambiguity is resolved to an a priori less expected structure (e.g., based on the statistics of the lexical context). We find that comprehenders rapidly adapt their syntactic expectations to converge towards the local statistics of novel environments. Specifically, repeated exposure to a priori unexpected structures can reduce, and even completely undo, their processing disadvantage (Experiment 1). The opposite is also observed: a priori expected structures become less expected (even eliciting garden paths) in environments where they are hardly ever observed (Experiment 2). Our findings suggest that, when changes in syntactic statistics are to be expected (e.g., when entering a novel environment), comprehends can rapidly adapt their expectations, thereby overcoming the processing disadvantage that mistaken expectations would otherwise cause. Our findings take a step towards unifying insights from research in expectation-based models of language processing, syntactic priming, and statistical learning.
When we read or listen to language, we are faced with the challenge of inferring intended messages from noisy input. This challenge is exacerbated by considerable variability between and within speakers. Focusing on syntactic processing (parsing), we test the hypothesis that language comprehenders rapidly adapt to the syntactic statistics of novel linguistic environments (e.g., speakers or genres). Two self-paced reading experiments investigate changes in readers’ syntactic expectations based on repeated exposure to sentences with temporary syntactic ambiguities (so-called “garden path sentences”). These sentences typically lead to a clear expectation violation signature when the temporary ambiguity is resolved to an a priori less expected structure (e.g., based on the statistics of the lexical context). We find that comprehenders rapidly adapt their syntactic expectations to converge towards the local statistics of novel environments. Specifically, repeated exposure to a priori unexpected structures can reduce, and even completely undo, their processing disadvantage (Experiment 1). The opposite is also observed: a priori expected structures become less expected (even eliciting garden paths) in environments where they are hardly ever observed (Experiment 2). Our findings suggest that, when changes in syntactic statistics are to be expected (e.g., when entering a novel environment), comprehenders can rapidly adapt their expectations, thereby overcoming the processing disadvantage that mistaken expectations would otherwise cause. Our findings take a step towards unifying insights from research in expectation-based models of language processing, syntactic priming, and statistical learning.
Human language is characterized by variability in that the way in which language is used varies depending, for example, on facts about the identity of the speakers or author, the social context, and surrounding linguistic material. Variability poses formidable challenges to the systems underlying language comprehension, which are known to exploit statistical contingencies in the input to overcome the inherent noisiness of perception; nevertheless, we seem to comprehend language with apparent ease. How is this possible? Here we argue that we are able to comprehend language efficiently in part by continuously adapting to the statistics of novel linguistic situations. We argue further that adaptation specifically allows comprehenders’ expectations to converge towards the actual statistics of the linguistic input. Concretely, we show that readers can adjust their linguistic expectations in light of recent experience such that (a) previously difficult structures become easier to process, and, even more strikingly, (b) previously easy to process structures come to incur a processing cost.
Biological Psychology
Electrophysiological and behavioral evidence of gender differences in the modulation of distraction by the emotional context2008 •
2006 •
Relative clauses (RCs) have been studied extensively in language acquisition and adult processing. Studies show that both children and adults find object relatives harder than subject relatives [1, 2]. Despite the similar pattern, adult difficulty is taken to reflect the increased processing demands of object relatives [1] while child difficulty is often interpreted as evidence for children's lack of adult-like knowledge of the structure, attributed to under-developed syntax [2] or to the use of non-adult processing heuristics which are ...
Frontiers in Psychology
Syntactic Priming As a Test of Argument Structure: A Self-paced Reading ExperimentUsing data from a behavioral structural priming experiment, we test two competing theoretical approaches to argument structure, which attribute different configurations to (in)transitive structures. These approaches make different claims about the relationship between unergatives and transitive structures selecting either a DP complement or a small clause complement in structurally unambiguous sentences, thus making different predictions about priming relations between them. Using statistical tools that combine a factorial 6 × 6 within subjects ANOVA, a mixed effects ANCOVA and a linear mixed effects regression model, we report syntactic priming effects in comprehension, which suggest a stronger predictive contribution of a model that supports an interpretive semantics view of syntax, whereby syntactic structures do not necessarily reflect argument/event structure in semantically unambiguous configurations. They also contribute novel experimental evidence that correlate representational complexity with language processing in the mind and brain. Our study further upholds the validity of combining quantitative methods and theoretical approaches to linguistics for advancing our knowledge of syntactic phenomena.
Manuscript
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Running head: VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Word count: ~9,000
The role of verb repetition in cumulative structural priming in comprehension
Alex B. Finea,b*
T. Florian Jaegerb,c
a
The Hebrew University of Jerusalem, Israel
Department of Psychology
b
University of Rochester
Department of Brain & Cognitive Sciences
c
University of Rochester
Department of Computer Science
Department of Linguistics
*Please
address correspondence to:
Alex B. Fine
Department of Psychology
Hebrew University of Jerusalem
Mt. Scopus
Jerusalem 91905
Israel
abfine@gmail.com
1
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Abstract
Recently processed syntactic information is likely to play a fundamental role in
online sentence comprehension. For example, there is now a good deal of evidence that
the processing of a syntactic structure (the target) is facilitated if the same structure was
processed on the immediately preceding trial (the prime), a phenomenon known as
structural priming. However, compared to structural priming in production, structural
priming in comprehension remains relatively understudied. We investigate an aspect of
structural priming in comprehension that is comparatively well understood in production,
but has received little attention in comprehension: the cumulative effect of structural
primes on subsequently processed sentences. We further ask whether this effect is
modulated by lexical overlap between preceding primes and the target. In three selfpaced reading experiments, we find that structural priming effects in comprehension are
cumulative and of similar magnitude both with and without lexical overlap. We discuss
the relevance of our results to questions about the relationship between recent experience
and online language processing.
Keywords: psycholinguistics, structural priming, adaptation, sentence processing
2
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
The role of verb repetition in cumulative structural priming in
comprehension
When speakers have the choice between near meaning-equivalent syntactic
structures (e.g., The clown gave the boy a balloon vs. The clown gave a balloon to the
boy), the probability of producing either of these structures increases after exposure to a
prime sentence with the same structure (e.g., speakers are more likely to produce the first
example above if they just produced a sentence like Tom threw the dog a bone, Bock,
1986). This phenomenon, known as structural priming, has played an important role in
research on language production (for reviews and theoretical proposals see Chang, Dell,
& Bock, 2006; Jaeger & Snider, 2013; Pickering & Garrod, 2012; Reitter, Keller, &
Moore, 2011). Structural priming in production has also been influential as a tool to
probe the nature of linguistic representations (for review, see Pickering & Ferreira, 2008).
Compared to production, structural priming in syntactic comprehension has
received considerably little attention. In sentence comprehension, structural priming
refers to the facilitated comprehension of a structure following exposure to that structure.
In a comprehensive review of the field, Pickering and Ferreira (2008) commented on the
striking sparsity of studies that investigated structural priming in comprehension in ways
parallel to production. Since then the field has seen a flurry of studies that have followed
the call for action. Recent work on priming has shown, for example, that garden-path
effects—where the underlined material in temporarily ambiguous relative clause (RC)
sentences such as (1a) takes longer to read than an unambiguous RC baseline (1b) or a
3
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
much more frequently-occurring temporarily ambiguous main verb (MV; 1c)—are
reduced when sentences such as (1a) are read immediately following a sentence with the
same structure, such as (2) (Ledoux, Traxler, & Swaab, 2007; Noppeney & Price, 2004;
Tooley, Traxler, & Swaab, 2009; Traxler & Tooley, 2008).
1. The soldiers…
a. Ambig. RC: …warned about the dangers conducted the midnight raid
b. Unambig .RC: ..who were warned about the dangers conducted the
midnight raid.
c. MV: …warned about the dangers before the midnight raid.
2. The workers warned about the wages decided to file complaints.
Similar trial-to-trial facilitation effects have been observed for a number of different
syntactic structures (e.g., sentences with modifier-goal ambiguities such as (3), where in
the box is parsed as either a modifier of or goal-PP for peanuts, depending on the prime,
Traxler, 2008; priming of early closure readings in sentences such as (4), where the
school is parsed as a subject NP rather than the object of left, Noppeney & Price, 2004;
Traxler, 2014; attachment ambiguities, where subjects can be primed to attach the
prepositional phrase in (5) to the man or the girl, Branigan, Pickering, & McLean, 2005).
3. The vendor tossed the peanuts in the box into the crowd during the game.
4. After the headmaster had left the school deteriorated rapidly.
5. The man saw the girl with the telescope.
4
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Structural priming in comprehension has now been demonstrated for both reading
(most of the work cited above, e.g., Ledoux et al., 2007; Tooley et al., 2009; Traxler,
2008) and auditory sentence processing (Arai & Mazuka, 2013; Arai, van Gompel, &
Scheepers, 2007; Kamide, 2012; Scheepers & Crocker, 2004; Thothathiri & Snedeker,
2008a, 2008b). The latter studies have also provided evidence that the facilitatory effect
of trial-by-trial structural priming stems at least partly from increased expectations for the
prime structure during processing of the target sentence (see also Fine & Jaeger, 2013).
Still, much remains to be understood about structural priming in comprehension. Here
we focus on the relation between trial-to-trial effects of exposure to a prime structure and
the cumulative effects on comprehension across many primes. Little is known about the
latter, although it has played an important role in research on production. Below, we
begin by introducing cumulative structural priming and its relevance to accounts of
structural priming. Then we introduce the property of cumulative structural priming we
focus on in this paper: the role of verb overlap between prime and target. As we discuss
below, verb overlap has figured prominently in research on trial-to-trial structural
priming in both production and comprehension, because it has been taken to speak to the
nature of the mechanisms involved in structural priming (for an overview, see Hartsuiker,
Bernolet, Schoonbaert, Speybroeck, & Vanderelst, 2008).
Trial-to-trial vs. cumulative structural priming
Research on structural priming in production has found that exposure to multiple
primes has a cumulative effect (e.g., Bock, Dell, Chang, & Onishi, 2007; Bock & Griffin,
2000; Branigan, Pickering, Stewart, & McLean, 2000; Hartsuiker, et al., 2008). For
5
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
example, the more often participants were exposed to double object (or prepositional
dative) primes during an exposure phase, the more likely they were to produce that
structure in a sentence completion task following the exposure phase (Kaschak, 2007).
The finding that priming is cumulative has played an important role in
distinguishing between accounts of syntactic priming in production. Cumulative effects
are not expected if priming is purely due to transient activation boosts (e.g., Branigan,
Pickering, & Cleland, 1999). Cumulativity has thus been taken to argue that structural
priming in production is at least in part due to learning (Jaeger & Snider, 2013; Kaschak,
Kutta, & Coyle, 2012; Kaschak, Kutta, & Jones, 2011; Kaschak, Kutta, & Schatschneider,
2011).
If structural priming in production and comprehension rely on the exact same
mechanisms (as hypothesized by, e.g., Chang et al., 2006; Tooley & Bock, 2014), we
would therefore expect similar cumulative priming effects in syntactic comprehension
(for a review of arguments for such comparative work, see Pickering & Ferreira, 2008;
Tooley & Traxler, 2010). In particular, if structural priming in comprehension is a side
effect of implicit learning during language processing (Chang et al., 2006; Fine & Jaeger,
2013; Jaeger & Snider, 2013; Luka & Barsalou, 2005), rather than transient activation
(Pickering, Branigan, & McLean, 2002), cumulativity is expected for comprehension, too.
A few studies have begun to address this question (Fine, Jaeger, Farmer, & Qian,
2013; Kaschak & Glenberg, 2004; Long & Prat, 2008; Wells, Christiansen, Race,
Acheson, & MacDonald, 2009). In a seminal paper, Wells et al. (2009) show that
repeated exposure to object-extracted relative clauses can, over the course of several days,
diminish the processing cost associated with this famously difficult-to-process structure,
6
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
highlighting the role of linguistic experience in sentence comprehension. In a similar
vein, Long and Prat (2008) exposed participants to temporarily ambiguous relative
clauses (RCs), as in (1), above, where warned about the dangers can be temporarily
interpreted as the matrix verb of the sentence.
Sentences such as (1a) are known to elicit garden-path effects at the
disambiguation point (…conducted…, e.g., Ferreira & Henderson, 1990). The magnitude
of the garden path effect can be ameliorated when the matrix subject biases toward the
intended RC reading rather than against it (like The evidence vs. The witness, respectively,
in (6); McRae, Ferretti, & Amyote, 1997; McRae, Spivey-Knowlton, & Tanenhaus, 1998;
Pearlmutter & MacDonald, 1995; Tabossi, Spivey-Knowlton, McRae, & Tanenhaus,
1994).
6. The {evidence, witness} examined by the lawyer was unreliable.
Long and Prat (2008) found that, after multiple days of exposure to sentences in
which animate subjects (the witness) always occurred with the a priori expected MV
continuation and inanimate subjects (the evidence) always occurred with RCs, subjects
who did not initially exploit the plausibility information (i.e., low-span readers) came to
benefit from this lexical cue.
The studies by Wells et al. (2009) and Long and Prat (2008) demonstrate that
long-term training through massive exposure to syntactic structures that are otherwise
hard to process can facilitate processing of these structures. These works leave open,
however, whether similar cumulative facilitation effects can be observed on a shorter
7
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
time scale, for instance, within an experiment. Such rapid cumulativity has been
observed for structural priming in production (Jaeger & Snider, 2013; Kaschak et al.,
2012; Kaschak et al., 2011).
Evidence that cumulative facilitation of a priori difficult structures can occur after
similarly brief exposure comes from more recent studies (Craycraft, 2014; Farmer, Fine,
Yan, Cheimariou, & Jaeger, 2014; Fine et al., 2013; Fine, Qian, Jaeger, & Jacobs, 2010;
Kaschak & Glenberg, 2004). Fine et al. (2013) used materials similar to those used by
Long and Prat (2008) to investigate whether repeated exposure to RCs within a single
experimental session would lead to cumulative reductions in the processing cost
associated with that structure. This is indeed what was observed: after exposure to around
25 temporarily ambiguous stimuli of the type shown above in (1) (a total of 36 critical
items were presented), subjects exhibited no detectable garden path effect (see also Fine
et al., 2010 for another type of temporary ambiguity). Parallel to evidence from trial-bytrial structural priming, this cumulative effect of exposure within a single experimental
session has also been found to be reflected in anticipatory eye-movements (Kamide,
2012), suggesting that it is at least in part due to adaptation of expectations about the
relative frequency of the primed structure (for a proposal along these lines as well as
further evidence in favor of it, see Fine et al., 2013).
Here we have two goals. If confirmed, evidence of rapid cumulative structural
priming would suggest that structural priming in comprehension is cumulative at time
scales similar to those observed for structural priming in production. Our first goal
therefore is to replicate these effects. Second, we aim to contribute to a better
understanding of how these cumulative effects relate to trial-to-trial priming effects in
8
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
syntactic comprehension (e.g., Branigan et al., 2005; Noppeney & Price, 2004;
Thothathiri & Snedeker, 2008a; Tooley et al., 2009; Traxler & Tooley, 2008; Traxler,
2008). To this end, we investigate the role of verb overlap between the prime and target
(or, in the case of cumulative priming, verb repetition across critical items), which has
been the target of many recent investigations of trial-to-trial priming in syntactic
comprehension, but has not previously been investigated in cumulative structural priming
in comprehension.1
Verb repetition in structural priming
The emerging body of work on cumulative structural priming (none of the studies
cited above use this term) has largely focused on questions about implicit learning—
whether experience with a structure improves subsequent processing (Long & Prat, 2008;
Wells et al., 2009); whether experiment-specific distributional information can be
extracted from the input (Fine et al., 2013; Kamide, 2012); and whether mere exposure
can lead subjects to acquire entirely novel grammatical constructions (Kaschak &
Glenberg, 2004). By asking whether cumulative structural priming is sensitive to verb
repetition, the current study provides an initial step towards unifying previous work on
cumulative priming on the one hand and previous work on trial-to-trial structural
priming—where the effect of verb repetition has been a central focus—on the other.
Some previous studies suggest that priming in comprehension occurs only when
prime and target sentences share the same verb (Arai et al., 2007; Ledoux et al., 2007;
Tooley et al., 2009; Traxler & Pickering, 2005; Traxler & Tooley, 2008), while others
1
What we refer to as verb overlap or verb repetition is also sometimes referred to as lexical overlap, though
the focus is usually on verbs, or, more generally, the head of the phrase being primed.
9
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
suggest that priming can occur with or without repeated verbs (Arai & Mazuka, 2013;
Fine et al., 2013, 2010; Kim, Carbary, & Tanenhaus, 2013; Thothathiri & Snedeker,
2008b; Traxler, 2008). Still other studies suggest that, while structural priming in
comprehension does not require repeated verbs across prime and target, priming is
stronger when this situation holds (Tooley & Bock, 2014; Traxler, Tooley, & Pickering,
2014; Traxler, 2014). This magnification of priming effects in the presence of verb
overlap between the prime and target is known as the “lexical boost” effect, and was first
observed in language production (e.g., Hartsuiker et al., 2008; Pickering & Branigan,
1998).
The relationship between trial-to-trial and cumulative priming in comprehension,
and whether they are sensitive to the same representational information, is a pertinent
question in light of recent formal models of priming in production suggesting that an
adequate model of structural priming likely needs to include a rapidly-decaying
component (captured by "transient activation" in some models, cf. Branigan, Pickering,
& Cleland, 1999; Pickering & Branigan, 1998; Pickering & Garrod, 2004, or by
declarative memory in others, Chang et al., 2006) and a longer-lasting, error-sensitive
implicit learning mechanism (Chang et al., 2006; Jaeger & Snider, 2013). (For explicit
discussions of the need for two mechanisms, see Chang et al., 2006; Hartsuiker et al.,
2008; Reitter et al., 2011.).
In contrast to dual-mechanism accounts of priming, some researchers have
suggested that the short-lived lexical boost may also follow from more general principles
of learning or expectation adaptation. Specifically, Jaeger and Snider (2013) suggest that
the rapidly-decaying lexical boost observed in production priming may reflect the fact
10
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
that content words follow a “bursty” distribution in spontaneously-produced spoken and
written language (Heller, Pierrehumbert, & Rapp, under revision; Katz, 1996): rather
than occurring with a uniform probability across linguistic environments, words tend to
occur in “bursts” of high local probability. In comprehension, listeners may therefore
expect verbs (and their associated structures) to occur in “clusters” (for some direct
evidence in support of this, see Myslin & Levy, under review), which would lead to a
rapidly-decaying lexical boost. However, the same reasoning may also lead to the
prediction that verb repetition will have longer-lasting effects, assuming that subjects
(implicitly) reason that the entire experimental environment constitutes a “cluster”,
within which verb-structure pairs are likely to be repeated. The goal of the current study
is not to adjudicate between dual-mechanism and single-mechanism, expectation
adaptation accounts of priming (nor do our experiments constitute an adequate test for
distinguishing these two very broad frameworks). We mention these divergent
perspectives here simply to highlight the rich array of previous work that is relevant to
our current question. We return to these theoretical questions in the discussion.
In the current set of experiments, if verb repetition was present in an experiment,
items with repeated verbs were separated by, on average, roughly 20 intervening
sentences. Therefore, if dual-mechanism models are correct, and structural priming is
mediated by the same mechanism in comprehension and production, then the lexical
boost should rapidly decay in comprehension as well, and cumulative structural priming
should not be influenced by verb repetition. We have shown in our previous work that
cumulative structural priming for such materials can be observed without verb overlap
11
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
(Fine et al., 2013), but it remains an open question whether this effect interacts with verb
repetition.
The current study
In three self-paced reading experiments we examine the effect of repeated verbs
on cumulative structural priming. Subjects read ambiguous and unambiguous sentences
like (1a) and (1b), repeated below as (7).
7. The experienced soldiers/…
a. Ambig. RC: …warned about the dangers/ conducted the midnight/ raid
b. Unambig .RC: ..who were/ warned about the dangers/ conducted the
midnight/ raid.
Sentences like (7a) are temporarily ambiguous because warned about the dangers
can be parsed either as the main verb (MV) phrase of the sentence, or as a relative clause
(RC) modifying soldiers. Previous work has shown that reading times (RTs) during the
disambiguating region (i.e., the point in the sentence where the MV reading is ruled out
in (7a), underlined above) are significantly higher for temporarily ambiguous RC
sentences relative to unambiguous RC sentences like (7b). We will refer to this
difference as the ambiguity effect (also known as the garden-path effect, discussed above,
cf. Frazier, 1987).
In the experiments reported in this paper, the ambiguity was always resolved as an
RC. Therefore, cumulative experience in the experimental environment should lead to
cumulative priming, which we quantify as the incremental and cumulative reduction in
12
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
the ambiguity effect.2 In Experiments 1 and 2, we present subjects with sentences like
(7), and measure the change in the ambiguity effect as subjects progress through the
experiment. In Experiment 1, verbs were systematically repeated across the experiment;
in Experiment 2, a different verb was used in each critical sentence. In Experiment 3, we
completely eliminated repetition of all content words, including verbs, nouns, adjectives,
and adverbs. We quantitatively compare the cumulative priming effect across
experiments to ask what effect verb repetition has on cumulative structural priming in
comprehension.
Experiment 1: Cumulative structural priming with verb repetition
Subjects
88 subjects were recruited via Amazon’s crowdsourcing platform Mechanical Turk. Only
subjects with US IP addresses were allowed to participate. Subjects were self-reported
native speakers of English, and only subjects with at least a 95% approval rating from
previous jobs were included. There is by now a wealth of evidence suggesting that
psycholinguistic experiments administered over the web replicate results obtained in labbased experiments, even with online measures such as self-paced reading (e.g., Demberg,
2013; Enochson & Culbertson, 2015; Keller, Gunasekharan, Mayo, & Corley, 2009;
Munro et al., 2010).
In the current study, we assume that both unambiguous and ambiguous RCs “count” as RCs to the subject,
and that the cumulative effect of observing both is what leads to cumulative syntactic priming. It is
possible that temporarily ambiguous RCs would lead to stronger cumulative priming effects (following an
account along the lines of that proposed by Kaschak & Glenberg, 2004). We leave this question to future
work.
2
13
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Materials
Critical items were constructed from sentence pairs like (7a)-(7b). Eight different verbs
giving rise to the MV/RC ambiguity (watched, washed, taught, served, called, warned,
dropped, pushed) were repeated 5 times to yield 40 critical items with different
lexicalizations apart from the verbs (i.e., different NPs, adjectives, etc.). Ambiguity was
counter-balanced across two experimental lists.
In addition, each list contained the same 80 fillers. Filler sentences featured a
variety of syntactic structures and, crucially, were constructed so as not to include verbs
that give rise to the MV/RC ambiguity (e.g., All the undergraduates in the class had
trouble keeping up; The foreign delegates arrived at the embassy surrounded by security
guards). All critical items and fillers, for both experiments, are provided in Appendix A.
Items and fillers were presented in the exact same pseudo-random order in both
lists. The pseudo-random order adhered to the constraint that two critical items not occur
on consecutive trials, that the same condition not be repeated more than three times in a
row, and that the experiment begin and end with filler trials. Moreover, these lists were
constructed such that the 8 verbs used to create the critical items were presented in 5
consecutive blocks, and no verb was repeated within a block. Thus critical items
containing the same verb were separated, on average, by 8 critical items (SD = 3) or by
an average of 23 critical and filler items (SD = 8). Next, the order of the two lists
counterbalancing ambiguity was reversed, yielding two presentation orders. This step is
important in that it decreases the correlation between item identity and presentation order.
This is particularly crucial for experiments on cumulative priming, since it guarantees
14
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
that effects of amount of cumulative exposure to a given syntactic structure is not
confounded with the identity of the items that contained that structure.
Procedure
Subjects read sentences in a self-paced moving window display (Just, Carpenter,
& Woolley, 1982). At the beginning of each trial, the sentence appeared on the screen
with all non-space characters replaced by a dash. Subjects pressed the space bar using
their dominant hand to view each consecutive word in the sentence. Durations between
space bar presses were recorded. At each press of the space bar, the currently-viewed
word reverted to dashes as the next word was converted to letters. A yes/no
comprehension question followed all experimental and filler sentences, with the correct
answer to half of all comprehension questions being “yes”. Subjects required at most 30
minutes to complete the experiment.
Results
Data coding and exclusions. Following common practice in the analysis of selfpaced reading data, RTs less than 100 ms or greater than 2000 ms were excluded. This
resulted in .4% data loss.
Next, we excluded trials on which subjects answered comprehension questions
incorrectly, resulting in an additional 7% data loss, averaging across critical items and
fillers (within critical items only, there was 9% data loss, with 10% of questions for
ambiguous items being answered incorrectly and 8% for unambiguous items being
answered incorrectly; this difference was significant, replicating MacDonald et al., 1992).
For all experiments, we excluded subjects with a comprehension question accuracy rate
below 80%. No subjects in Experiment 1 were affected by this criterion. After
15
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
exclusions, the average by-participants comprehension accuracy was 93% (SD=3). None
of the results for any of the experiments presented here depend on the exclusion of
incorrectly-answered items (after we summarize the reading time results, we also present
an analysis of comprehension question accuracy across all three experiments).
We then performed a residualization step intended to remove the effect of word
length on log reading times (RTs).3 Word length is a strong predictor of word-by-word
RTs, and it is standard in analyzing RTs to regress RTs onto word length and then to use
the residuals of this model as a dependent measure (Ferreira & Clifton, 1986). We
therefore regressed log RTs in the entire data set (excluding only practice trials) against
word length (in characters) using linear mixed effects regression (Baayen, Davidson, &
Bates, 2008). The model also included a by-subject random slope for word length. These
random effects allow the model to discount mean differences in RTs across subjects as
well as variable sensitivity to the effect of word length across subjects. The residuals of
this model yield residual log RTs, which will serve as the dependent measure in all
analyses reported throughout the manuscript.
Analysis. Although the sentences were read one word at a time, for the purposes
of analysis we followed standard practice and segmented sentences into regions indicated
by the forward slashes in (7) above. We designate these regions the subject, the
relativizer (only present in unambiguous sentences, e.g., who were in (7b)), the
3
We log-transformed raw reading times because this transformation (a) allows RTs to more closely satisfy
the assumption of normality, and (b) led to (mildly) better model fits than non-transformed reading times.
The results reported in this paper do not depend on this decision, though the non-critical ambiguity effect
during the ambiguous region of Experiment 2 is significant when log-transformed RTs are used, but only
marginally significant when non-transformed RTs are used.
16
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
ambiguous region, the disambiguating region, and the final word. These regions are the
same used by MacDonald, Just, and Carpenter (1992).
We begin by plotting by-region mean residual log RTs (i.e., residual log RTs
averaged across words within each region) in Figure (1). This figure serves to
demonstrate that, before looking for evidence of structural priming, we have replicated
the ambiguity effects found in previous work. Consistent with previous research, Figure
(1) shows larger RTs for ambiguous (dark, solid lines) relative to unambiguous (light,
dashed lines) sentences. This effect surfaced in the ambiguous and disambiguating
regions, and spilled over onto the final word of the sentence (cf. Table 1). This replicates
MacDonald et al. (1992), who also found significant ambiguity effects at these three
regions.
*** Insert Figure 1 here ***
Cumulative structural priming is predicted to surface as a change in the ambiguity
effect over the course of the experiment. In the following, we thus focus on the
disambiguating region, in which the ambiguity effect is observed (indeed, as shown in
Table 1 below, the effects of interest only surfaced in this region, as predicted). We
regressed mean residual log RTs during the disambiguating region onto ambiguity
(ambiguous vs. unambiguous), item order (coded 1-40), and the interaction between the
two. In order to control for task adaptation (i.e., an overall speed-up in RTs across all
17
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
regions due to increasing familiarity with the self-paced reading paradigm), we included
a main effect of log-transformed stimulus order. Stimulus order differs from item order in
that it is an index of when an item was presented relative to both items and fillers.
Stimulus order therefore captures how long (i.e., for how many trials) subjects have been
doing the experiment, providing a measure of adaptation to the self-paced reading task;
item order captures the number of RCs observed by subjects at a given point in the
experiment, providing a measure of cumulative priming (specifically, the two-way
interaction between ambiguity and item order). Here and in all experiments, all predictors
were mean-centered to reduce collinearity with higher-order interactions (remaining fixed
effects correlation rs<.2 except for high collinearity between the main effects of item
order and log stimulus order, r = -.8).4 In this and all other analyses, the maximal random
effects structure justified by the design was used (in Experiment 1, this corresponded to
random intercepts for subject and item, as well as by-subject and by-item random slopes
for ambiguity, item order, and their two-way interaction).
Results: Reading times on the disambiguating region. There was a main effect
of ambiguity: RTs during the disambiguating region were significantly slower in
ambiguous sentences relative to unambiguous sentences (=.03, SE =.003, p < .001).5
The main effect of log stimulus order also reached significance ( = -.1, SE = .01, p
4
In the analyses for all three experiments, item order and log stimulus order have highly correlated effects
(i.e., are highly collinear). We leave both in the models reported below since the two are theoretically
distinct and often account for unique variance. Neither of the two effects is critical to our argument. By
including stimulus order, we rule out the possibility that the critical effect of cumulative priming (i.e., a
two-way interaction between ambiguity and item order) is driven by actual changes in subjects’
expectations for the RC structure, rather than simply an overall decrease in RTs. In our previous work, we
have discussed this issue in detail, ruling out this possibility both statistically and experimentally (see Fine
et al., 2010; 2013). Collinearity between these two predictors does not affect the coefficients or standard
errors of any other predictors.
5
All significance levels are based on the t-distribution, under the assumption that, for data sets of this size,
t-values with absolute values greater than 1.96 are significant at =.05.
18
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
< .001), indicating that the disambiguating region was read faster with increasing
exposure to the task. The main effect of item order did not reach significance ( = -.002,
SE = .001 p = .2). Crucially, and as predicted, there was a significant two-way
interaction between ambiguity and log item order (= -.001, SE = .0002, p < .001): the
ambiguity effect significantly decreased as subjects observed more and more RCs over
the course of the experiment. To graphically demonstrate the effect of item order over
and above stimulus order, we residualized raw RTs against both word length and stimulus
order. Plotting these residual (log) RTs more clearly reveals the underlying cause of the
change of the ambiguity effect—a diminished processing cost for ambiguous RCs
relative to unambiguous RCs. This effect is visualized in Figure 2.
*** Insert Figure 2 here ***
The two-way interaction between ambiguity and log item order was observed
only during the disambiguating region and (marginally) during the final word, likely
reflecting “spillover”. The results of the model described above, fit separately to each
sentence region, are summarized in Table 1.6
*** Insert Table 1 here ***
6
For the relativizer region there is no coefficient for ambiguity, or interactions involving ambiguity, since
this region was only present in unambiguous stimuli.
19
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Results: comprehension question accuracy. A common concern in experiments
like this (and, indeed, a concern raised by our reviewers) is that self-paced reading can be
tedious. It is possible that subjects stop paying attention to the sentences they are reading
in order to more quickly finish the task. This is sometimes thought to cause increasingly
faster reading times, in particular, on difficult sentences such as ambiguous RCs. If this
intuition is confirmed, this would constitute a confound. If the reading time effects we
reported above are indeed driven by decreasing attention, we should see decreases in the
rate of correctly-answered comprehension questions.
To address this question, we used mixed-effects logistic regression to analyze
comprehension question accuracy on critical trials. To brevity’s sake, we present the
analyses of comprehension question accuracy for all three experiments here. The
comprehension question accuracy for ambiguous and unambiguous items in all three
experiments is summarized in Table 2.
*** Insert Table 2 here ***
We regressed comprehension question accuracy onto ambiguity, item order, and the twoway interaction between these two predictors. The main effect of ambiguity went in the
same direction for all three experiments (significant for Experiment 1 and 2): subjects
were more likely to incorrectly answer questions after ambiguous RCs. This replicates
previous work (e.g., MacDonald et al., 1992) in suggesting that sentences with
ambiguous RCs are indeed harder to process. The main effect of item order was not
significant in Experiments 1 and 2 and positive and marginally significant in Experiment
20
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
3 (β = 0.02, SE = 0.008, p = .09), suggesting that subjects in Experiment 3 became
slightly more accurate, overall, as the experiment progressed. The crucial two-way
interaction between ambiguity and item order was in the same direction in all three
experiments (significant in Experiments 1 and 3; marginally significant in Experiment 2,
p = .1): subjects became better at answering comprehension questions for ambiguous
items relative to unambiguous items (confirmed by simple effect analyses, which found
this improvement to be marginally significant for Experiment 1 and significant for
Experiment 3); accuracy on unambiguous RCs remained unchanged throughout the
experiments (all simple effect analyses, p > .2).
The analyses of comprehension question accuracy thus provide no evidence for
decreasing attention due to boredom or fatigue. Rather, we see—in Experiments 1 and
3—an improvement in comprehension of ambiguous RCs, consistent with the idea that
exposure facilitates processing.
Discussion
The results of Experiment 1 show that, as subjects read RC sentences, the
processing advantage conferred by early disambiguating information (who were…, in the
example above) on reading times during the later “disambiguating” region diminishes.
We interpret this as evidence of cumulative structural priming. This finding replicates
previous work on cumulative priming in syntactic comprehension (Farmer et al., 2014;
Fine et al., 2013, 2010; Kamide, 2012; Kaschak & Glenberg, 2004; see also Long & Prat,
2008; Wells et al., 2009). In Experiment 2, we ask whether this effect depends on
repeated verbs among the critical items.
21
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Experiment 2: Cumulative Structural priming without verb repetition
Experiment 2 involved exactly the same materials and procedure as Experiment 1,
but without verb repetition. Eight of the critical items in this experiment were included in
Experiment 1. The remaining 32 critical items were created by replacing the verb in one
of the critical items from Experiment 1 with a near-synonym (e.g., changing pushed to
shoved in An impatient shopper shoved through the doors complained to the manager).
The critical items from Experiment 2 are included in Appendix A.
The same 80 filler sentences were used here as in Experiment 1. Experiment 2
employed the exact same method to create four experimental lists (and two pseudorandom presentation orders) as Experiment 1, although items and fillers were not
presented in the same exact order across Experiments 1 and 2.
Subjects
80 subjects saw 40 critical items in one of the four possible lists. Subjects were recruited
via Mechanical Turk according to the same procedure described for Experiment 1.
Subjects required at most 30 minutes to complete the experiment.
Results
Data coding and exclusions. As in Experiment 1, RTs less than 100ms or
greater than 2000ms were excluded before computing residual log RTs. This resulted in
1% data loss. Next, we excluded trials on which subjects answered comprehension
questions incorrectly (see Table 2). Two subjects with low accuracy rates on the
comprehension questions (<80%) were excluded, leaving 78 subjects for the analyses
reported below. After excluding these subjects, the average by-participants
comprehension accuracy was 92% (SD=3).
22
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Analysis. By-region mean residual log RTs (i.e., averaging across words within
each region) are plotted in Figure 3. The same broad patterns evident in Figure 1 are
present in Figure 3, with significant ambiguity effects at the ambiguous, disambiguating,
and final word regions, replicating both Experiment 1 and MacDonald et al. (1992) (cf.
Table 3).
*** Insert Figure 3 here ***
We then conducted the same analysis as in Experiment 1 to examine changes in
the ambiguity effect over the course of the experiment. As in the analysis of the data
from Experiment 1, we centered all predictors. Collinearity was generally low (r < .2),
with the exception of high collinearity between the main effects of item order and log
stimulus order (r = -.9). The maximal random effects structure justified by the design that
would converge corresponded to random intercepts for subject and item, as well as bysubject and by-item random slopes for ambiguity, item order, and their interaction. We
found main effects of ambiguity ( = .03, SE = .004, p < .001) , item order ( = -.005, SE
= .001, p < .001), and log stimulus order ( = -.07, SE = .01, p < .001): as in Experiment
1, residual log RTs during the disambiguating region were overall higher for ambiguous
relative to unambiguous items, and decreased with cumulative exposure to RC structures.
Finally, there was a significant two-way interaction between ambiguity and log item
order ( = -.001, SE = .0001, p < .01). We plot this effect in Figure 4, again using lengthand stimulus order-residualized log RTs.
23
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
*** Insert Figure 4 here ***
The two-way interaction was observed in the disambiguating region. As in
Experiment 1, the effect reached only marginal significance in the final word region (β =
-.001, SE = .0004, p = .1). The results of this model, fit separately to the data from each
sentence region, are summarized in Table 3.
*** Insert Table 3 here ***
Comparison of Experiments 1 and 2. Qualitatively speaking, Experiment 2
replicated Experiment 1: we observed significant main effects—of similar magnitudes—
of ambiguity and item order, as well as a two-way interaction between these two
variables, in both experiments. Our results so far suggest that verb repetition is not
required for cumulative priming in syntactic comprehension. This echoes some previous
trial-to-trial priming results (Arai & Mazuka, 2013; Kim et al., 2013; Thothathiri &
Snedeker, 2008a, 2008b; Traxler, 2008). Next, we ask whether there is a quantitative
difference in the priming effect across these experiments that is attributable to verb
repetition, i.e., a lexical boost effect in cumulative priming in comprehension.
First, we computed residual log RTs using the aggregated data from Experiments
1 and 2 in the manner described above (analyzing raw RTs returns identical results). This
residualization step was repeated on the aggregated data so that our analysis can detect
experiment-specific effects (all effects replicate if RTs are residualized separately for
each experiment and then combined). Next, we analyzed residual log RTs against the full
24
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
factorial design of ambiguity, item order, and experiment (Experiment 1 or Experiment 2),
as well as a main effect of log stimulus order. We included random intercepts for
subjects and items, as well as by-subject and -item random slopes for ambiguity, item
order, and their two-way interaction).
Replicating the separate analyses of Experiments 1 and 2, the aggregate analysis
of the disambiguation region revealed significant main effects of ambiguity (β = .03, SE
= .003, p < .001), item order (β = -.003, SE = .001, p < .01), log stimulus order (β = -.1,
SE = .001, p < .001) and the interaction between ambiguity and item order (β = -.001, SE
= .0001, p < .001).
Most germane to the question of whether the effects observed across the two
experiments differed in any way, neither the main effect of experiment nor any of the
interactions in which it participated reached significance on the disambiguating region
(p’s > .4). The results for all other sentence regions are summarized in Table 4.
*** Insert Table 4 here ***
With few exceptions, the results at all other sentence regions are as expected, with
significant effects of stimulus order across all regions (and effects of item order at the
disambiguating and final word regions), ambiguity effects at the ambiguous,
disambiguating, and final word regions, and a two-way interaction between ambiguity
and item order during the disambiguating region and the final word region.
Two effects were not expected. First, there was a main effect of experiment
during the ambiguous region: during this region, residual log RTs were slightly lower for
25
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Experiment 1 than Experiment 2. This is somewhat surprising since the material during
these regions was the same across experiments, except for the verb (the effect of
experiment remains significant even when the verb is excluded from this region, though it
is possible that such a lexical priming effect could spill over into the rest of the
ambiguous region.7). A second possible explanation for this effect is that the different
semantic properties of the verbs across the two experiments generate different
expectations during the ambiguous region, leading to differences in overall RTs in the
ambiguous region. 8 There was also an unexpected effect of the three-way interaction
during the subject region. We have no explanation for this effect.
Discussion
We manipulated verb repetition across Experiments 1 and 2 and found that the
cumulative priming effect is of a quantitatively indistinguishable magnitude in both
experiments. The results are consistent with the hypothesis that priming in
comprehension is guided by the same mechanism as priming in production, and that this
mechanism is comprised of a long-lasting implicit learning mechanism that is not
sensitive to repeated lexical material and a short-lived advantage for repeated verbs, due
either to declarative memory (Chang et al., 2006) or transient boosts in the activation of
lemma information (Branigan et al., 1999; Malhotra, 2009).
Experiment 3: Cumulative structural priming without any lexical repetition
7
We thank an anonymous reviewer for this suggestion.
It seems unlikely that the main effect of experiment is responsible for the lack of a significant three-way
interaction at the disambiguating region, though it is possible that the different thematic relations, tied to
different verbs, across the two experiments worked against and thereby masked an underlying difference in
the adaptation effects across the two experiments.
8
26
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Experiment 3 involved a modified version of the design and materials used in
Experiments 1 and 2. Both Experiments 1 and 2 unintentionally contained repetition of
lexical items apart from the verb (see Appendix A). Twenty non-verb content words
were repeated on average two times (SD = 0.3), averaging across all sentence regions.
This affected 25 out of 40 items (that is, 60% of items contained one content word, e.g.,
hot, that also occurred in another critical item). This non-verb lexical repetition was
identical across Experiments 1 and 2. Such non-verb repetition across Experiments 1 and
2 is compatible with our goal to assess the effect of verb overlap on cumulative structural
priming (most of previous work has focused on the role of verb overlap, or, more
generally, lexical repetition of the head of the structure).
However, there is also some evidence from production that non-head repetition
can also affect structural priming. For instance, Cleland and Pickering (2003) find
stronger priming effects in production when prime and target share a subject noun phrase
(see also Snider, 2008). Between our Experiments 1 and 2, seven out of 40 critical items
contained subject NPs that overlapped lexically with the subject NPs of other items. It is
thus possible that we failed to find a difference between Experiments 1 and 2 because
repeated nouns, adjectives, and adverbs strengthened the priming effects in both
experiments, making it harder to detect an effect of verb repetition. To address this
possibility, care was taken in Experiment 3 to eliminate all repeated content words across
critical and filler items. Additional motivation for Experiment 3 comes from the fact that
the comparison of Experiments 1 and 2 yielded a null effect with regard to verb overlap.
One must always exercise caution when interpreting null effects, and replicating the
effect constitutes good practice as an initial step in interpreting it.
27
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Subjects
Experiment 3 employed the exact same procedure as Experiments 1 and 2 above.
80 subjects participated in this study, and three were removed due to low (<80%)
accuracy rates on the comprehension questions.
Results
Data coding and exclusions. We excluded trials with abnormally high
(>2000ms) and low (<100ms) RTs, as well as trials on which subjects answered
comprehension questions incorrectly. After exclusions, mean by-subject accuracy on the
comprehension questions was 93% (SD = 3), averaging across critical items and fillers.
Analysis. By-region mean residual log RTs (i.e., averaging across words within
each region) are plotted in Figure 5. The same broad patterns evident in Figures 1 and 3
are present in Figure 5, with significant ambiguity effects at the ambiguous,
disambiguating, and final word regions, replicating Experiments 1 and 2 and MacDonald
et al. (1992) (cf. Table 5). One unexpected effect of ambiguity was observed during the
subject region—with unambiguous items being read more slowly. We comment on this
below.
*** Insert Figure 5 here ***
We analyzed residual log RTs during the disambiguating region using the same
modeling procedure employed in the analysis of the data from Experiments 1 and 2. The
model included random intercepts for subjects and items, as well as by-subject and byitem random slopes for ambiguity, item order, and their interaction. This experiment
28
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
replicated Experiments 1 and 2: there were significant main effects of ambiguity (β = .03,
SE = .004 p < .001) and log stimulus order (β = -.1, SE = .02 p < .001). The main effect
of item order did not reach significance (β = .001, SE = .001, p = .6). Crucially, we
observed the predicted two-way interaction between ambiguity and item order (β = -.001,
SE = .0001, p < .001). The two-way interaction is shown in Figure 6. Once again, we
visualize the cumulative priming effect using length- and stimulus order-residualized log
RTs
*** Insert Figure 6 here ***
We summarize the results of the regression model described above, fit separately
to the data from each sentence region, in Table 5.
*** Insert Table 5 here ***
The two-way interaction was observed only during the disambiguating region, but
not in the final word region (the latter was significant in Experiments 1 and marginally
significant in Experiment 2). The ambiguity effect was observed at the ambiguous region,
the disambiguating region, and the final word, replicating Experiments 1-2, as well as
MacDonald et al. (1992). We also observed an effect of log stimulus order across all
regions, reflecting overall speedup in RTs over the course of the experiment.
In Experiment 3 we also observed an unexpected “ambiguity effect”—slightly
greater RTs for unambiguous sentences—during the subject region. This effect was
29
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
surprising given that the material during the subject region was identical across
ambiguous and unambiguous sentences, which might explain the small magnitude of the
effect. We have no explanation for this effect.
Comparing Experiments 1 and 3. We next tested whether there were any
differences in the cumulative structural priming effects observed across Experiments 1
and 3. As in the comparison of the data from Experiments 1 and 2, we began by
computing residual log RTs using the aggregated data from Experiments 1 and 3. Next,
we regressed residual log RTs onto the full factorial design of ambiguity, item order, and
experiment (Experiment 1 or Experiment 2), as well as a main effect of log stimulus
order. We included the maximal random effects structure justified by the design (i.e.,
random intercepts for subjects and items, as well as by-subject and -item random slopes
for ambiguity, item order, and their two-way interaction).
The results of this comparison echo the results of the comparison, above, of
Experiments 1 and 2. We found—replicating the analyses of each of the separate
experiments—a significant main effect of log stimulus order (β = -.1, SE = .005, p
< .001), a significant main effect of ambiguity (β = .03, SE = .005, p < .001), and a
significant two-way interaction between ambiguity and item order (β = -.001, p < .001).
Moreover, the three-way interaction between ambiguity, item order, and experiment—
which captures the quantitative difference in the magnitude of the cumulative structural
priming effects across the two experiments—was not significant (p = .5). In other words,
even after completely removing repeated lexical items from the materials used in
Experiment 2, there is no evidence that cumulative syntactic priming in comprehension is
any weaker than in Experiment 1, in which verbs were systematically repeated (and other
30
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
content words were unsystematically repeated). The results of this model, fit separately
to each sentence region, are summarized in Table 6.
*** Insert Table 6 here ***
Echoing the results of Experiments 1 and 3, we observed a main effect of log
stimulus order across all sentence regions (with a significant main effect of item order
only during the final word), as well as a significant main effect of ambiguity during the
ambiguous, disambiguating, and final word regions. The two-way interaction between
ambiguity and item order was also observed at the disambiguating region. As in the
separate analysis of Experiment 3, we observed a significant main effect of ambiguity
during the subject region, such that unambiguous items were read more slowly during
this region. This effect was not observed during the subject region in the separate
analysis of Experiment 1, though the two-way interaction between experiment and
ambiguity during the subject region only approached significance (β = .003, SE = .01, p
= .19).
Finally, replicating the comparison of Experiments 1 and 2 above, we found the
same two somewhat surprising effects. First, there was a main effect of experiment
during the ambiguous region, such that reading times were slightly faster for Experiment
1 relative to Experiment 3 (again, the effect was driven by Experiment 1). As stated
above, this effect may be attributable to different semantic properties of the verbs across
the two experiments generating different expectations during the ambiguous region,
leading to differences in overall RTs in the ambiguous region (recall that the same verbs
31
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
were used in Experiments 2 and 3). Second, a hint of the unexpected three-way
interaction between ambiguity, item order, and experiment—observed in the comparison
of Experiments 1 and 2—was observed here (marginally significant at p = .1). That these
two effects replicate to some degree is not surprising, as a) both comparisons involve
Experiment 1 and b) Experiments 2 and 3 were mostly identical stimuli (the same
differences were not observed when comparing Experiments 2 and 3, p > .5).
Discussion
In Experiment 3 we sought to replicate the cumulative structural priming effects
observed in Experiments 1 and 2 after removing all repeated content words in the
materials of Experiments 1 and 2. By replicating the crucial effects of Experiments 1 and
2—main effects of ambiguity and log stimulus order, as well as a two-way interaction
between ambiguity and item order during the disambiguating region—the results of
Experiment 3 provide further evidence for cumulative structural priming in
comprehension that does not require verb repetition. In short, the results of Experiment
3—as well as the results of the comparison of Experiments 1 and 3—suggest that
cumulative syntactic priming in comprehension does not depend on lexical repetition. In
the General Discussion, we elaborate on the ramifications of our results.
General Discussion
This paper set out to examine whether cumulative priming during syntactic
comprehension is sensitive to verb repetition. In three self-paced reading experiments we
found clear evidence for cumulative structural priming but no evidence that this effect
depended on verb repetition. Specifically, independent of verb repetition, we found that
32
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
the ambiguity effect for MV/RC-ambiguous sentences significantly diminished—and
eventually was undone completely—as subjects accumulated experience with RCs.
Before discussing the theoretical implications of our results, two points bear brief
emphasis. The first is that a failure to observe an effect of verb repetition is a null effect,
and one must always exercise caution when interpreting such effects. That being said, it
is important to note that several previous papers have claimed that there is no structural
priming in comprehension without verb overlap (Arai et al., 2007; Ledoux et al., 2007;
Tooley et al., 2009; Traxler & Pickering, 2005; Traxler & Tooley, 2008). Given that we
observe strong cumulative priming, a “null effect” of a boost for verb overlap—when
priming is still observed both with and without overlap—is informative. Second, results
like those reported in this paper suggest that traditional sentence processing experiments
employing similar designs may be systematically underestimating the magnitude of
garden-path and other effects (see Fine et al., 2013; Jaeger, 2010 for further discussion).
The lexical boost and the mechanism(s) underlying structural priming
As we discussed in the introduction, the current study builds on previous work on
cumulative priming in comprehension (Fine et al., 2013; Kamide, 2012; Kaschak &
Glenberg, 2004; Long & Prat, 2008; Wells et al., 2009), and extends this work by
examining cumulative priming along a dimension previously overlooked in that literature
(verb repetition) that has been discussed extensively in the trial-to-trial priming literature.
While there is some amount of equivocation with respect to this question in trialto-trial priming in comprehension, the emerging consensus seems to be that structural
priming does not require repeated verbs, but can be strengthened (or “boosted”) by it
(Tooley & Bock, 2014; Traxler et al., 2014; Traxler, 2014). In syntactic production, a
33
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
much more substantial body of work on structural priming has also shown evidence for a
lexical boost (Hartsuiker et al., 2008; Pickering & Branigan, 1998). Interestingly, in
syntactic production, while priming effects for lexically-independent syntactic structures
seem to be long-lived, the lexical boost appears to rapidly decay (but see Coyle &
Kaschak, 2008). This poses one of the central empirical challenges for models of
priming in syntactic production, with so-called transient activation accounts readily
accounting for short-lived as well as lexically-specific priming effects (e.g., Pickering &
Branigan, 1998) while not accounting for longer-lasting learning-based effects, and
implicit learning accounts readily accounting for long-lasting effects of abstract structural
priming (Chang et al., 2006) while not straightforwardly predicting the lexical boost. In
response to this challenge, current models of structural priming in production often
propose a mechanism comprised of two components—one suitable to capturing shortlived changes in the processing of a given structure, and another capturing longer-lasting
changes. For example Reitter et al. (2011) propose an ACT-R based model of priming in
production in which distributional patterns over syntactic structures are implicitly learned,
while lemma- (or verb-) based information is subject to rapid power-law decay. Chang et
al. (2006) take a different tack and argue that long-lived effects of syntactic priming are
driven by an error-driven implicit learning mechanism (implemented via the backpropagation learning algorithm in a connectionist architecture), while the lexical boost
effect is mediated by explicit (declarative) memory for specific lexical items that is
known to very rapidly decay (cf. Chang et al., 2006, p. 256 for discussion; as well as
Chang, Janciauskas, & Fitz, 2012, for more recent work that develops this hypothesis).
34
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Given the state of the field in structural priming in production, in order to
understand whether priming in comprehension and production share a similar or identical
underlying cause, it is crucial to fill in the empirical picture and ask whether the effect of
lexical repetition behaves the same over short and longer timescales in comprehension, or
whether the lexical boost in comprehension is similarly short-lived. The results of the
current experiments, taken together with previous work showing evidence for the lexical
boost in comprehension (Tooley & Bock, 2014; Traxler et al., 2014; Traxler, 2014)
suggest that the lexical boost in structural priming in comprehension is short-lived, as has
been shown in production (Hartsuiker et al., 2008).
On the face of it, then, our results seem to provide further support for the
mechanistic parity between structural priming in comprehension and production, as
previously argued by Chang et al., 2006 and Tooley & Bock, 2014. In the introduction
we briefly mentioned a different account of structural priming—related to recent work on
expectation adaptation (Fine et al., 2013; Jaeger & Snider, 2013; Kleinschmidt & Jaeger,
2015)—that provides an alternative perspective on the effect of verb repetition in priming.
Specifically, if comprehenders are sensitive to (a) the “bursty” distribution that lexical
items (including verbs) follow in natural language, and (b) the correlation between verbs
and syntactic structures (Trueswell, 1996), it is possible that comprehenders will expect
specific verb-structure pairs to be repeated within a given linguistic environment, which
would surface as a lexical boost effect in cumulative priming in comprehension. While
our results do not support this hypothesis, there are at least two plausible explanations as
to why we may not have observed an effect of verb repetition that follow naturally from
some recent work on expectation adaptation (Craycraft, 2014; Fine et al., 2013; Jaeger &
35
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Snider, 2013; Kleinschmidt & Jaeger, 2015; Myslín & Levy, n.d.). First, it is possible that
the materials in our experiments were sufficiently thematically and semantically
unrelated (as is the case in most lists of sentence processing stimuli) that subjects did not
(implicitly) infer that specific verb-structure pairs were likely to be repeated over the
course of the experiment. If this is correct, then experimental materials such as those we
used here could be modified in order to suggest the existence of a common underlying
“topic” that might lead to longer-lasting effects of verb repetition. We leave this to future
work.
Second, it is possible that the lack of a lexical boost effect in Experiment 1
(relative to Experiments 2 and 3) is due to the fact that verbs in our experiments are not
informative with respect to syntactic structure: in all Experiments reported above,
sentences with MV/RC ambiguities were always resolved towards the RC interpretation,
regardless of the verb. Even if participants come into our experiments with the
expectation that verbs are informative about the syntactic structures that follow them (as
the works cited above strongly suggest, especially Trueswell, 1996), they might quickly
revise this expectation in light of the statistical properties of the experiment. Future work
can address this question by testing whether verb-specific structural priming is observed
in environments (experiments) in which verbs remain informative about the structures
that follow them. For instance, if one set of MV/RC-ambiguous verbs always occurs with
RCs and another set always occurs with MVs for the first portion of an experiment, then
processing RCs in a subsequent portion of the experiment should be facilitated for
sentences including those verbs that always occurred with RCs earlier in the experiment.
For preliminary tests of this hypothesis on the current data, please see Appendix B.
36
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Trial-to-trial and cumulative structural priming: bridging the gap
Partially independent of the question of whether the same mechanism underlies
structural priming in comprehension and production, we believe the current results –
along with other recent research on cumulative structural priming in comprehension—
take an important step in bridging the gap between what we have referred to as trial-totrial priming in comprehension on the one hand and cumulative priming in
comprehension on the other. The connection between trial-to-trial and cumulative
priming effects has received little attention in previous work, even when it was arguably
relevant (though see Fine et al., 2013 for discussion). Here, we highlight two questions in
which the two paradigms can mutually inform each other.
One such question relates to the persistence of structural priming effects in
comprehension. For example, in a recent paper Tooley and colleagues (Tooley, Swaab,
Boudewyn, Zirnstein, & Traxler, 2014) examine trial-to-trial priming in comprehension
when multiple (manipulated to range from 0-3) unrelated sentences intervene between
primes and targets with overlapping verbs. Tooley and colleagues found that structural
priming persisted over intervening trials. This replicates earlier work by Fine et al. (2013;
Experiment 1, cf. Figure 5), who find cumulative priming effects for the same structure
(sentences with temporarily ambiguous relative clauses) in a paradigm in which RCs are
separated, on average, by 4 intervening sentences (see also Fine et al., 2010, as well as
the current results). The current results further suggest that persistent cumulative priming
is also observed in the absence of verb overlap (see also Craycraft, 2014; Farmer et al.,
2014; Fine et al., 2013).
37
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Another area in which cumulative priming paradigms like the one employed here
can inform research on trial-to-trial priming relates to the role of cue informativity.
Recent studies on cumulative structural priming have found that comprehenders can
exploit the informativity or reliability of, say, lexical or contextual cues to syntactic
structure in order to learn the environment-specific statistics of syntactic constructions
(Craycraft, 2014; Myslin & Levy, submitted; for related work in production, see Coyle &
Kaschak, 2008). This seems to contrast with studies on trial-to-trial priming that have
addressed related questions (Kim & Mauner, 2006; Traxler & Tooley, 2008). For
example, Traxler and Tooley (2008) found that repeated subject NPs across prime-target
pairs did not lead to significant priming, and interpret this lack of an effect as evidence
against “strategic prediction” accounts of syntactic priming. One possibility, however, is
that the effects of repeated subject NPs are too small to be observed in trial-to-trial
priming, or that they only emerge after participants have learned that, in the current
environment, the lexical identity of subject NPs is informative about upcoming syntactic
structures. Indeed, recent evidence from a series of studies on cumulative structural
priming in comprehension lends credence to the latter idea. Craycraft (2014) found that
participants could both learn and exploit the fact that, in their experiment, subject NPs
were correlated with specific syntactic structures. We consider this an interesting venue
for future research. For example, it is possible to re-analyze studies on trial-to-trial
priming, using statistical procedures like those employed in the current study. This can
shed light on how trial-to-trial priming changes throughout the course of the experiment
(for examples, see Arai & Mazuka, 2013; Jaeger & Snider, 2013).
38
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
In conclusion, we found no difference in the magnitude of the strength of
cumulative structural priming during comprehension depending on verb repetition. With
respect to theoretical debates informing most previous work on syntactic priming, these
results provide further support for the mechanistic parity of priming in comprehension
and production, with longer-lived learning effects operating over abstract structural
information and short-lived boosts in the activation of lexical-structural information (cf.
Tooley & Bock, 2014). That being said, this study is likely only one of many steps that
must be taken to fully understand the relationship between online processing and recent
linguistic experience, and the mechanism or mechanisms that mediate this relationship.
References
Arai, M., & Mazuka, R. (2013). The development of Japanese passive syntax as indexed
by structural priming in comprehension. Quarterly Journal of Experimental
Psychology, 3–8. doi:10.1080/17470218.2013.790454
Arai, M., van Gompel, R. P. G., & Scheepers, C. (2007). Priming ditransitive structures
in comprehension. Cognitive Psychology, 54(3), 218–50.
doi:10.1016/j.cogpsych.2006.07.001
Baayen, R. H., Davidson, D. J., & Bates, D. (2008). Mixed-effects modeling with crossed
random effects for subjects and items. Journal of Memory and Language, 59, 390–
412.
Bock, K. (1986). Syntactic persistence in language production. Cognitive Psychology,
18(3), 355–387. doi:10.1016/0010-0285(86)90004-6
Bock, K., Dell, G. S., Chang, F., & Onishi, K. (2007). Persistent structural priming from
language comprehension to language production. Cognition, 104(3), 437–458.
Bock, K., & Griffin, Z. M. (2000). The Persistence of Structural Priming: Transient
Activation or Implicit Learning? Journal of Experimental Psychology: General,
129(2), 177–192. doi:10.10371/D096-3445.129.2.1T/The
39
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Branigan, H. P., Pickering, M. J., & Cleland, A. A. (1999). Syntactic priming in written
production: Evidence for rapid decay. Psychonomic Bulletin and Review, 6(4), 635–
640.
Branigan, H. P., Pickering, M. J., & McLean, J. F. (2005). Priming prepositional-phrase
attachment during comprehension. Journal of Experimental Psychology: Learning,
Memory & Cognition, 31, 468–481.
Branigan, H. P., Pickering, M. J., Stewart, A. J., & McLean, J. F. (2000). Syntactic
priming in spoken production: Linguistic and temporal interference. Memory &
Cognition, 28, 1297–1302.
Chang, F., Dell, G. S., & Bock, K. (2006). Becoming syntactic. Psychological Review,
113(2), 234–72. doi:10.1037/0033-295X.113.2.234
Chang, F., Janciauskas, M., & Fitz, H. (2012). Language adaptation and learning: Getting
explicit about implicit learning. Language and Linguistics Compass, 6(5), 259–278.
doi:10.1002/lnc3.337
Coyle, J. M., & Kaschak, M. P. (2008). Patterns of experience with verbs affect longterm cumulative structural priming. Psychonomic Bulletin & Review, 15(5), 967–70.
doi:10.3758/PBR.15.5.967
Craycraft, N. (2014). Effect of reliability on syntactic adaptation. University of Rochester.
Demberg, V. (2013). Integration Costs on Auxiliaries? A self-paced reading study using
WebExp. In M. Knauff, N. Pauen, Sebanz, & I. Wachsmuth (Eds.), Proceedings of
the 35th annual meeting of the Cognitive Science Society. Berlin.
Enochson, K., & Culbertson, J. (2015). Collecting psycholinguistic response time data
using Amazon Mechanical Turk. PloS One.
Farmer, T. A., Fine, A. B., Yan, S., Cheimariou, S., & Jaeger, T. F. (2014). Syntactic
expectation adaptation in the eye-movement record. In Proceedings of the 36th
Annual Meeting of the Cognitive Science Society (pp. 2181–2186). Austin, TX:
Cognitive Science Society.
Farmer, T. A., Monaghan, P., Misyak, J. B., & Christiansen, M. H. (2011). Phonological
typicality influences sentence processing in predictive contexts: A reply to Staub et
al. (2009). Journal of Experimental Psychology: Learning, Memory, and Cognition,
37, 1318–1325.
Ferreira, F., & Clifton, C. E. (1986). The independence of syntactic processing. Journal
of Memory and Language, 25, 348–368.
40
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Ferreira, F., & Henderson, J. M. (1990). The use of verb information in syntactic parsing:
A comparison of evidence from eye movements and word-by-word self-paced
reading. Journal of Experimental Psychology: Learning, Memory & Cognition, 16,
555–568.
Fine, A. B., & Jaeger, T. F. (2013). Evidence for implicit learning in syntactic
comprehension. Cognitive Science, 37(3), 578–591.
Fine, A. B., Jaeger, T. F., Farmer, T. A., & Qian, T. (2013). Rapid Expectation
Adaptation during Syntactic Comprehension. PLoS ONE, 8(10), 1–18.
Fine, A. B., Qian, T., Jaeger, T. F., & Jacobs, R. A. (2010). Syntactic Adaptation in
Language Comprehension. In Proceedings of the 2010 Workshop on Cognitive
Modeling and Computational Linguistics (pp. 18–26). Uppsala, Sweden:
Association for Computational Linguistics.
Frazier, L., & Fodor, J. D. (1979). On comprehending sentences: syntactic parsing
strategies. University of Massachusetts.
Garnsey, S., Pearlmutter, N., Myers, E., & Lotocky, M. (1997). The contributions of verb
bias and plausibility to the comprehension of temporarily ambiguous sentences.
Journal of Memory and Language, 37, 58–93.
Hartsuiker, R. J., Bernolet, S., Schoonbaert, S., Speybroeck, S., & Vanderelst, D. (2008).
Syntactic priming persists while the lexical boost decays: Evidence from written and
spoken dialogue. Journal of Memory and Language, 58(2), 214–238.
doi:10.1016/j.jml.2007.07.003
Heller, J. R., Pierrehumbert, J. B., & Rapp, D. N. (n.d.). How fine-grained is discourse
predictability? Word burstiness affects lexical access.
Jaeger, T. F. (2010). Redundancy and reduction: speakers manage syntactic information
density. Cognitive Psychology, 61(1), 23–62. doi:10.1016/j.cogpsych.2010.02.002
Jaeger, T. F., & Snider, N. E. (2013). Alignment as a consequence of expectation
adaptation: syntactic priming is affected by the prime ’ s prediction error given both
prior and recent experience. Cognition, 127(1), 57–83.
Just, M. A., Carpenter, P. A., & Woolley, J. D. (1982). Paradigms and processes in
reading comprehension. Journal of Experimental Psychology: General, 111(2),
228–38. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/6213735
Kamide, Y. (2012). Learning individual talkers’ structural preferences. Cognition, 124(1),
66–71. doi:10.1016/j.cognition.2012.03.001
41
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Kaschak, M. P. (2007). Long-term structural priming affects subsequent patterns of
language production. Memory & Cognition, 35, 925–937.
Kaschak, M. P., & Glenberg, A. M. (2004). This construction needs learned. Journal of
Experimental Psychology. General, 133(3), 450–67. doi:10.1037/00963445.133.3.450
Kaschak, M. P., Kutta, T. J., & Coyle, J. M. (2012). Long and short term cumulative
structural priming effects. Language and Cognitive Processes.
Kaschak, M. P., Kutta, T. J., & Jones, J. L. (2011). Structural priming as implicit
learning: Cumulative priming effects and individual differences. Psychonomic
Bulletin and Review, 18, 1133–1139.
Kaschak, M. P., Kutta, T. J., & Schatschneider, C. (2011). Long-term cumulative
structural priming persists for (at least) one week. Memory & Cognition, 39, 381–
388.
Katz, S. M. (1996). Distribution of content words andphrases in text and language
modelling. Natural Language Engineering, 2, 15–59.
Keller, F., Gunasekharan, S., Mayo, N., & Corley, M. M. B. (2009). Timing Accuracy of
Web Experiments: A Case Study Using the WebExp Software Package. Behavior
Research Methods, 41, 1–12.
Kim, C., Carbary, K., & Tanenhaus, M. K. (2013). Syntactic Priming without Lexical
Overlap in Reading Comprehension. Language and Speech, 57(2).
Kim, S., & Mauner, G. (2006). Priming vs. contingency learning accounts of structural
priming effects in comprehension. In Proceedings of the 19th Annual Meeting of the
CUNY Sentence Processing Conference.
Kleinschmidt, D., & Jaeger, T. F. (2011). A Bayesian Belief Updating Model of Phonetic
Recalibration and Selective Adaptation. In Proceedings of the 2nd Workshop on
Cognitive Modeling and Computational Linguistics (pp. 10–19). Portland, OR:
Association for Computational Linguistics.
Kleinschmidt, D., & Jaeger, T. F. (2015). Robust speech perception: Recognize the
familiar, generalize to the similar, and adapt to the novel. Psychological Review.
Ledoux, K., Traxler, M. J., & Swaab, T. Y. (2007). Syntactic priming in comprehension:
evidence from event-related potentials. Psychological Science, 18(2), 135–43.
doi:10.1111/j.1467-9280.2007.01863.x
42
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Levin, B. (1993). English Verb Classes and Alternations: A Preliminary Investigation.
Chicago, IL: University of Chicago Press.
Long, D. L., & Prat, C. S. (2008). Individual differences in syntactic ambiguity
resolution: Readers vary in their use of plausibility information. Memory &
Cognition, 36(2), 375–391.
Luka, B., & Barsalou, L. (2005). Structural facilitation: Mere exposure effects for
grammatical acceptability as evidence for syntactic priming in comprehension.
Journal of Memory and Language, 52(3), 436–459. doi:10.1016/j.jml.2005.01.013
MacDonald, M. C. (1999). Distributional information in language comprehension,
production, and acquisition: Three puzzles and a moral. In The Emergence of
Language (pp. 177–196). Mahwah, NJ: Erlbaum.
MacDonald, M. C. (2013). How language production shapes language form and
comprehension. Frontiers in Psychology, 4, 1–16.
MacDonald, M. C., Just, M. A., & Carpenter, P. A. (1992). Memory Constraints on the
Processing Syntactic Ambiguity. Cognitive Psychology, 98, 56–98.
Malhotra, G. (2009). Dynamics of structural priming. University of Edinburgh.
McRae, K., Ferretti, T., & Amyote, L. (1997). Thematic roles as verb-specific concepts.
Language and Cognitive Processes, 12, 137–176.
McRae, K., Spivey-Knowlton, M. J., & Tanenhaus, M. K. (1998). Modeling the influence
of thematic fit (and other constraints) in on-line sentence comprehension. Journal of
Memory and Language, 38, 283–312.
Munro, Bethard, Kuperman, V., Lai, Melnick, R., Potts, C., … Tily. (2010).
Crowdsourcing and language studies: The new generation of linguistic data. In
Proceedings of NAACL 2010.
Myslín, M., & Levy, R. (n.d.). Comprehension priming as rational expectation for
repetition.
Noppeney, U., & Price, C. J. (2004). An FMRI study of syntactic adaptation. Journal of
Cognitive Neuroscience, 16(4), 702–13. doi:10.1162/089892904323057399
Pearlmutter, N. J., & MacDonald, M. C. (1995). Individual differences and probabilistic
constraints in syntactic ambiguity resolution. Journal of Memory and Language, 34,
521–542.
43
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Pickering, M. J., & Branigan, H. P. (1998). The representation of verbs: Evidence from
syntactic priming in language production. Journal of Memory and Language, 39,
633–651.
Pickering, M. J., & Ferreira, V. S. (2008). Structural priming: A critical review.
Psychological Bulletin, 134(3), 427–459.
Pickering, M. J., Branigan, H., & McLean, J. (2002). Constituent structure is formulated
in one stage. Journal of Memory and Language, 46(3), 586–605.
Pickering, M. J., & Garrod, S. (2004). Toward a mechanistic psychology of dialogue.
Behavioral and Brain Sciences, 27(2), 169–226. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/15595235
Pickering, M. J., & Garrod, S. (2013). An integrated theory of language production and
comprehension. Behavioral and Brain Sciences, 36(4), 329–347.
Reitter, D., Keller, F., & Moore, J. D. (2011). A computational cognitive model of
syntactic priming. Cognitive Science, 35(4), 587–637. doi:10.1111/j.15516709.2010.01165.x
Roland, D., Dick, F., & Elman, J. L. (2007). Frequency of basic English grammatical
structures: A corpus analysis. Journal of Memory and Language, 57, 348–379.
doi:10.1016/j.jml.2007.03.002
Scheepers, C., & Crocker, M. W. (2004). Constituent order priming from listening to
comprehension: A visual-world study. In M. Carreiras & J. Clifton, C. (Eds.), The
online study of sentence comprehension: Eyetracking, ERPs, and beyond (pp. 167–
185). New York: Psychology Press.
Snider, N. (2008). An exemplar model of syntactic priming. Stanford University.
Tabossi, P., Spivey-Knowlton, M. J., McRae, K., & Tanenhaus, M. K. (1994). Semantic
effects on syntactic ambiguity resolution: Evidence for a constraint-based resolution
process. In Attention & Performance XV. Cambridge, MA: MIT Press.
Tanenhaus, M. K., Spivey-Knowlton, M. J., Eberhard, K., & Sedivy, J. (1995).
Integration of visual and linguistic information in spoken language comprehension.
Science, 268, 1632–1634.
Thothathiri, M., & Snedeker, J. (2008a). Give and take: syntactic priming during spoken
language comprehension. Cognition, 108(1), 51–68.
doi:10.1016/j.cognition.2007.12.012
44
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Thothathiri, M., & Snedeker, J. (2008b). ThothathiriSnedeker08b. Journal of Memory
and Language, 58, 188–213.
Tooley, K. M., & Bock, K. (2014). On the parity of structural persistence in language
production and comprehension. Cognition, 132(2), 101–36.
doi:10.1016/j.cognition.2014.04.002
Tooley, K. M., Swaab, T., Boudewyn, M., Zirnstein, M., & Traxler, M. (2014). Evidence
for priming across intervening sentences during on-line sentence comprehension.
Language, Cognition and Neuroscience, 29(3), 289–311.
Tooley, K. M., & Traxler, M. J. (2010). Syntactic Priming Effects in Comprehension: A
Critical Review. Language and Linguistics Compass, 4(10), 925–937.
doi:10.1111/j.1749-818X.2010.00249.x
Tooley, K. M., Traxler, M. J., & Swaab, T. Y. (2009). Electrophysiological and
behavioral evidence of syntactic priming in sentence comprehension. Journal of
Experimental Psychology: Learning, Memory & Cognition, 35(1), 19–45.
Traxler, M. J. (2008). Lexically independent priming in online sentence comprehension.
Psychonomic Bulletin & Review, 15(1), 149–155. doi:10.3758/PBR.15.1.149
Traxler, M. J. (2014). Priming of Early Closure: Evidence for the Lexical Boost during
Sentence Comprehension. Language, Cognition and Neuroscience, 30(4), 1–13.
Traxler, M. J., & Pickering, M. J. (2005). Syntactic priming in comprehension: Evidence
from eye- movements. In Proceedings of the 18th annual CUNY conference on
human sentence processing. Tucson, AZ.
Traxler, M. J., & Tooley, K. M. (2008). Priming in sentence comprehension: Strategic or
syntactic? Language and Cognitive Processes, 23(5), 609–645.
doi:10.1080/01690960701639898
Traxler, M. J., Tooley, K. M., & Pickering, M. J. (2014). Syntactic priming during
sentence comprehension: Evidence for the lexical boost. Journal of Experimental
Psychology: Learning, Memory & Cognition, 40(4), 905–918.
Trueswell, J. C., Tanenhaus, M. K., & Kello, C. (1993). Verb-specific constraints in
sentence processing: Separating effects of lexical preference from garden-paths.
Journal of Experimental Psychology: Learning, Memory and Cognition, 19(3), 528–
553.
Wells, J. B., Christiansen, M. H., Race, D. S., Acheson, D. J., & MacDonald, M. C.
(2009). Experience and sentence processing: statistical learning and relative clause
45
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Item Order
β
t
β
t
β
t
Final word
region
Disambiguating
region
Relativizer
Subject
Predictor
Ambiguous
comprehension. Cognitive Psychology, 58(2), 250–71.
doi:10.1016/j.cogpsych.2008.08.002
β
t
β
t
0.002
1.826
-0.003
-2.245
-0.002
-1.494
-0.002
-1.846
-0.004
-2.247
Log stimulus order
-0.130
-8.906
-0.077
-4.512
-0.108
-8.564
-0.103
-7.215
-0.078
-2.792
Ambiguity (=ambiguous)
-0.006
-1.511
NA
0.019
6.103
0.026
7.516
0.021
3.210
Ambiguity : Item order
-0.001
-1.790
NA
-0.001
-0.590
-0.001
-4.620
-0.001
-1.898
Table 1. Coefficients and t-values for each predictor (rows), at each sentence region
columns), in Experiment 1. For data sets of this size, t-values with an absolute value of
1.96 or greater are significant at p≤.05. Significant effects are in bold.
46
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Experiment 1
Experiment 2
Experiment 3
Ambiguous
10%
12%
10%
Unambiguous
8%
8%
8%
Table 2. Comprehension question accuracy rates (after subject exclusions) for
ambiguous and unambiguous items in Experiments 1-3.
47
β
β
t
t
β
t
β
Final word
region
Disambiguating
region
Relativizer
Subject
Predictor
Ambiguous
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
t
β
t
Item Order
-0.002
-1.415
-0.002
-1.612
-0.003
-2.698
-0.005
-3.810
-0.005
-2.757
Log stimulus order
-0.078
-4.636
-0.083
-3.658
-0.085
-5.635
-0.071
-5.185
-0.083
-3.818
Ambiguity (= ambiguous)
-0.002
-0.569
NA
0.012
2.218
0.026
5.848
0.022
3.149
0.001
1.068
NA
0.001
1.039
-0.001
-2.405
-0.001
-1.768
Ambiguity : Item order
Table 3. Coefficients and t-values for each predictor (rows), at each sentence region
columns), in Experiment 2. For data sets of this size, t-values with an absolute value of
1.96 or greater are significant at p≤.05. Significant effects are in bold.
48
β
t
β
t
β
t
Final word
region
Disambiguating
region
Relativizer
Subject
Predictor
Ambiguous
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
β
t
β
t
0.004
0.622
0.007
1.124
-0.013
-2.828
0.003
0.482
0.017
1.106
Item Order
-0.001
-0.169
-0.002
-1.555
-0.003
-2.957
-0.003
-3.951
-0.004
-3.266
Log stimulus order
-0.997
-8.769
-0.094
-6.599
-0.096
-9.319
-0.086
-8.310
-0.085
-4.808
Ambiguity ( = ambiguous)
-0.003
-1.472
0.015
5.082
0.026
9.430
0.022
4.579
Experiment : Item Order
0.001
0.586
0.001
0.435
0.001
0.454
0.001
0.902
Experiment : Ambiguity
-0.002
-0.698
NA
0.004
1.214
-0.001
-0.074
-0.001
-0.009
Ambiguity : Item Order
-0.001
-0.756
NA
0.001
0.626
-0.001
-5.048
-0.001
-2.412
Experiment : Ambig. : Item Order
-0.001
-2.007
NA
-0.003
-1.207
-0.001
-0.785
-0.001
-0.128
Experiment (= repeated)
NA
-0.001
-0.190
Table 4. Coefficients (and t-values) for each predictor (rows), at each sentence region columns),
for the combined data from Experiments 1 and 2. For data sets of this size, t-values with an
absolute value of 1.96 or greater are significant at p≤.05. Significant effects are in bold.
49
Item Order
β
t
β
t
β
Final word
region
Disambiguating
region
Relativizer
Subject
Predictor
Ambiguous
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
t
β
t
β
t
0.001
0.667
-0.001
-0.957
-0.001
-0.041
0.001
0.719
-0.003
-1.232
Log stimulus order
-0.086
-5.028
-0.071
-3.375
-0.111
-5.942
-0.129
-6.985
-0.102
-3.405
Ambiguity (=ambiguous)
-0.012
-3.287
NA
0.099
2.185
0.027
6.232
0.021
4.114
Ambiguity : Item Order
0.001
0.283
NA
-0.001
-0.177
-0.001
-3.327
-0.001
-0.075
Table 5. Coefficients and t-values for each predictor (rows), at each sentence region columns), in
Experiment 3. For data sets of this size, t-values with an absolute value of 1.96 or greater are
significant at p≤.05. Significant effects are in bold.
50
β
β
t
β
t
β
t
Final word
region
Disambiguating
region
Relativizer
Subject
Predictor
Ambiguous
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
β
t
t
Experiment (= 1)
0.006
0.893
-0.001
-1.838
-0.016
-3.596
0.003
0.472
0.029
1.887
Item Order
0.002
1.872
-0.002
-1.576
-0.001
-1.317
-0.001
-1.275
-0.003
-2.170
Log stimulus order
-0.110
-9.753
-0.077
-4.519
-0.103
-9.151
-0.106
-9.267
-0.091
-4.250
Ambiguity (=ambiguous)
-0.009
-3.654
0.014
5.346
0.026
9.768
0.022
5.162
Experiment : Item Order
0.001
1.417
-0.001
-1.340
-0.001
-1.625
-0.001
-0.399
Experiment : Ambiguity
-0.003
-1.326
NA
0.005
1.717
-0.001
-0.290
-0.001
-0.049
Ambiguity : Item Order
-0.001
-1.119
NA
-0.001
-0.553
-0.001
-4.692
-0.001
-1.469
0.001
1.658
NA
-0.001
-0.173
-0.001
-0.412
-0.001
-1.296
Experiment : Ambig. : Item Order
NA
-0.001
-1.509
Table 6. Coefficients and t-values for each predictor (rows), at each sentence region columns), for the
combined data from Experiments 1 and 3. For data sets of this size, t-values with an absolute value of
1.96 or greater are significant at p≤.05. Significant effects are in bold.
51
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 1. Mean residual log RTs at each sentence region in Experiment 1, plotted
separately for ambiguous (dark, solid lines) and unambiguous (light, dashed lines) items.
Error bars represent 95% confidence intervals on the means.
52
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 2. Change in residual log RTs from Experiment 1 during the disambiguating
region for ambiguous (circles and dark, solid line) and unambiguous (triangles and light,
dashed line) items as a function of item order. The plot reveals overall speed-ups in RTs
as the experiment progresses, as well as a decrease in the ambiguity effect, i.e.,
cumulative syntactic priming.
53
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 3. Mean residual log RTs at each sentence region in Experiment 2, plotted
separately for ambiguous (dark, solid lines) and unambiguous (light, dashed lines) items.
Error bars represent 95% confidence intervals on the means.
54
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 4. Change in residual log RTs from Experiment 2 during the disambiguating
region for ambiguous (squares and dark, solid line) and unambiguous (triangles and light,
dashed line) items as a function of item order. The plot reveals overall speed-ups in RTs
as the experiment progresses, as well as a decrease in the ambiguity effect, i.e.,
cumulative syntactic priming.
55
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 5. Mean residual log RTs at each sentence region in Experiment 3, plotted
separately for ambiguous (dark, solid lines) and unambiguous (light, dashed lines) items.
Error bars represent 95% confidence intervals on the means.
56
VERB REPETITION IN CUMULATIVE STRUCTURAL PRIMING
Figure 6. Change in residual log RTs from Experiment 3 during the disambiguating
region for ambiguous (squares and dark, solid line) and unambiguous (triangles and light,
dashed line) items as a function of item order. The plot reveals overall speed-ups in RTs
as the experiment progresses, as well as a decrease in the ambiguity effect, i.e.,
cumulative structural priming.
57
Supplemental Material - Integral
Click here to access/download
Integral Supplemental Material (made available to
referees)
Appendices.docx
RELATED PAPERS
Journal of Memory and Language
The activation of inappropriate analyses in garden-path sentences: Evidence from structural priming2006 •
Language, Cognition and Neuroscience
The use of syntax and information structure during language comprehension: Evidence from structural priming2019 •
2009 •
Language and Speech
Syntactic Priming without Lexical Overlap in Reading Comprehension2013 •
2003 •
2003 •
2003 •
2003 •
Psychological Science
Syntactic Priming in Comprehension: Evidence From Event-Related Potentials2007 •
Cognitive Science
Uncertainty and Expectation in Sentence Processing: Evidence From Subcategorization Distributions2015 •
Journal of Experimental Psychology: Learning, Memory, and Cognition
Priming Prepositional-Phrase Attachment During Comprehension2005 •
Proceedings of the 36th Annual Meeting of the Cognitive Science Society
Error-Driven Adaptation of Higher-Level Expectations During Reading2014 •
Proceedings of the 33rd Annual Conference of the Cognitive Science Society
Implicit Context-Specific Learning Leads to Rapid Shifts in Syntactic Expectations2011 •
Journal of Memory and Language
Gradiency and visual context in syntactic garden-paths2007 •
Journal of Experimental Psychology: Learning, Memory, and Cognition
Evidence for Self-Organized Sentence Processing: Digging-In Effects2004 •
Journal of Memory and Language
Distinguishing effects of structure and decay on attachment and repair: A cue-based parsing account of recovery from misanalyzed ambiguities2003 •
2008 •
Journal of Memory and Language
The Effects of Thematic Fit and Discourse Context on Syntactic Ambiguity Resolution2001 •
Journal of Memory and Language
The use of thematic role information in parsing: Syntactic processing autonomy revisited2003 •